A full investigation of the directional congestion in data envelopment analysis

One of interesting subjects in Data Envelopment Analysis (DEA) is estimation of congestion of Decision Making Units (DMUs). Congestion is evidenced when decreases (increases) in some inputs result in increases (decreases) in some outputs without worsening (improving) any other input/output. Most of the existing methods for measuring the congestion of DMUs utilize the traditional definition of congestion and assume that inputs and outputs change with the same proportion. Therefore, the important question that arises is whether congestion will occur or not if the decision maker (DM) increases or decreases the inputs dis-proportionally. This means that, the traditional definition of congestion in DEA may be unable to measure the congestion of units with multiple inputs and outputs. This paper focuses on the directional congestion and proposes methods for recognizing the directional congestion using DEA models. To do this, we consider two different scenarios: (i) just the input direction is available. (ii) none of the input and output directions are available. For each scenario, we propose a method consists in systems of inequalities or linear programming problems for estimation of the directional congestion. The validity of the proposed methods are demonstrated utilizing two numerical examples.